did you say advanced persistent threats? - welivesecurity did you say advanced persistent threats? 2

Download Did you say Advanced Persistent Threats? - WeLiveSecurity Did you say Advanced Persistent Threats? 2

Post on 28-May-2020

0 views

Category:

Documents

0 download

Embed Size (px)

TRANSCRIPT

  • Did you say Advanced Persistent Threats?

  • Did you say Advanced Persistent Threats?

    1

    Here we analyze four targeted attack tools with Taiwan and Vietnam in their sights - but somehow linked together - and the reason why they shouldn't be called ‘advanced’.

    Once in a while we get to spend time analyzing malicious code that is not as widespread or not as well-obfuscated as other threats we've encountered in the past. This article is about one such threat. We decided to spend some time on this analysis because of interesting strings in one of the components referring to Vietnam’s Central Post and Telecommunications Department. But before we delve into the topic lets first highlight some of the findings:

    • Entities in Taiwan and the Vietnam government are targeted • Observed attacker interaction • Evidence of an unidentified APT actor

    • Social engineering vector (no exploit code) with very credible documents

    • Bad criminals: typos in configuration, naive cryptographic implementation, weak code practices

    • Sophistication variability: from no obfuscation to hidden position independent code, XOR encryption, XTEA encryption, stand-alone re-usable components

    • Tailored infections: one threat doesn't persist, the other doesn't do anything before a reboot

    You can see in the above figure all the malware samples that this article will cover. The file received by the victim is always the dropper which we will cover shortly. Since they were carrying two different threats the dropper hashes are not the same but their functionality is equivalent: therefore it is summarized as a single threat and

    Figure 1: Targeted entities were located in Vietnam and Taiwan

    Figure 2: Analyzed threats

    http://www.virusradar.com/en/glossary/dropper-trojan

  • Did you say Advanced Persistent Threats?

    2

    considered a re-usable component in the attacker’s arsenal. We have investigated two ‘dropped’ threats, namely Agent.NJK and Terminator RAT – which also carries an embedded binary.

    Good ol’ social engineering

    As we noticed from our telemetry data, the malicious software reaches its target through spear-phishing campaigns. The first dropper we analyzed came from the webmail interface of a Vietnamese governmental institution. Using targeted emails allows more chance of succeeding in the attack by using a more personalized and convincing message. It also narrows the distribution of the malicious files, giving them a longer shelf life since there is less chance of their being found and analyzed by Anti-Virus (AV) companies.

    With knowledge of the characteristics of the first dropper, we were able to find a related piece of malware in our collection. As mentioned previously, they were carrying different threats but also had a different filenames

    Threat File name Translation

    Win32/TrojanProxy.Agent.NJK Bao cao ket qua.doc [137 spaces].exe

    Vietnamese for "report the results"

    Terminator RAT (Win32/Protux.NAR)

    檢驗報告.exe Chinese for "inspection report"

    The presence of all those spaces is used to push the ".exe" off the screen and out of sight of the victim. To further convince the user

    that the file is a normal Word document, the executable displays the icon of a Word document.

    Upon execution these droppers will decrypt their configuration parameters using a simple one-byte key XOR-based cipher best described with some python code below. This configuration is stored in the last 32 bytes of the last portable executable (PE) segment of the executable. Inside this configuration is a checksum, some offsets and lengths of internal resources along with other seemingly unused fields, as you will see in the struct pictured below. a hard- coded integer in the code is compared with the checksum in order to validate that configuration decryption worked. This checksum is

    Figure 3: Appearance of the files

    def xor _ decrypt(ciphertext, key): for i in range

    (len(ciphertext)): c = ciphertext[i] if c: if c != 0xff:

    c ^= key if (c and c != 0xff):

    ciphertext[i] = c return ciphertext

    struct hidden _ segment _ data { int checksum; char delimiter; char unused[3]; int pe _ file _ offset; int pe _ file _ size; char unused[4]; int doc _ file _ offset; int doc _ file _ size; char xorkey; char unused[2]; char last; };

    Listing 1: XOR-based cipher Listing 2: Hidden configuration

  • Did you say Advanced Persistent Threats?

    3

    the same in both cases. The offset and length pairs are used to extract files from inside itself into the filesystem.

    The dropper first drops the main malicious binary and then a Word document into the user's temporary folder. Both files are decrypted using the same simple XOR technique except that the malicious binary is prefixed with 5 bytes that are hard-coded in the dropper (MZ header), and then XOR'ed with another hardcoded one-byte key. We believe this is done to avoid being detected by some AV.

    First, after the extraction, the malicious binary will be executed by the dropper. The behavior of the analyzed binaries will be covered later. The dropper will then copy itself using a handle retrieved with GetModuleHandle. It will execute this fresh copy with some command line arguments in order to clean up after itself: namely, the current full path and filename of the dropper and the full path and filename of the dropped Word document. Finally, it will exit.

    For example this is what ends up being run:

    C:\Documents and settings\user\Local Settings\Temp\~hCb37.tmp\

    "C:\Documents and settings\user\Downloads\Bao cao ket qua.doc[137 spaces].exe"\

    "C:\Documents and settings\user\Local Settings\Temp\~hC29f.doc"

    Listing 3: Dropper executes the above

    Nature of the file Filename

    Malicious payload %TEMP%\~hCb58.tmp

    Word document %TEMP%\~hC29f.doc

    Copy of itself %TEMP%\~hCb37.tmp

    Table 1: Dropped files

    This same copy of the dropper, once executed with command-line arguments, has a different operation. It will first sleep for one second, leaving enough time for the original dropper execution to terminate. Then it will remove this original file and copy the decoy document (~hC29f.doc) in its place, keeping the proper .doc extension. Finally, a ShellExecuteW with the open operation is run on the newly copied document in order to open the proper editor registered for this file type.

    Figure 4: Dropper operation

  • Did you say Advanced Persistent Threats?

    4

    All this work is done to effectively simulate the result one would expect when double clicking on an innocuous Word document except that in this case malicious code was executed first.

    The combination of the spear-phishing, hiding the file's extension, a work-related file name and a Microsoft Word style icon can be pretty convincing for a user who had no proper security awareness training or without proper desktop hardening and protection against executables sent by email. The use of these simple techniques is well documented inside Mandiant's APT1 report. Notice that no software vulnerabilities are exploited by criminals in order to get their malware to run.

    In the dropper there are two different techniques used to hide calls: a function that essentially re-implements GetProcAddress, called

    Figure 5: Vietnam decoy document

    Figure 6: Taiwan decoy document

    http://intelreport.mandiant.com/Mandiant_APT1_Report.pdf

  • Did you say Advanced Persistent Threats?

    5

    with hardcoded plaintext strings, and legitimate GetProcAddress calls but using an obfuscated lpProcName (XOR 0x17 of every other two chars). Interestingly, most of the calls are not obfuscated. Again, it feels like iterative AV evasion hard at work.

    Aside from the fact that it seems easy to re-purpose, the dropper doesn't strike us as a particularly well written piece of code. There are notorious anti-patterns present in the codebase like a God object and some copy-and-paste programming (although to be fair this could be the result of compiler optimization).

    Win32/TrojanProxy.Agent.NJK

    The first dropped binary that we analyzed is what our engine detects as Win32/TrojanProxy.Agent.NJK. This is a Visual C++ trojan that communicates over HTTP with hard-coded Command and Control (C&C) servers. In the sample we analyzed, the three servers supported by the trojan configuration were in fact pointing to the same domain name vietnam.vnptnet.info, but using different ports (80, 443 and 5050).

    The malware will adjust its TCP timeout for HTTP requests to 15 minutes and then loop forever trying to contact the C&C domain via the three ports in configuration. An interesting fact about this threat is its lack of persistence, meaning that it will be executed only once and will not be relaunched if the system reboots. There is no obvious attempt at obfuscation and simply running strings on the binary reveals a great deal about the sample and its capabilities.

    In its attempt to contact the C&C the malware will send several pieces of information about the host in a GET request and use a specific User-Agent string. The user data is in a 105 bytes array, encoded in hexadecimal and sent in the path component of the GET request. It contains information such as: a string we believe is used to trac

Recommended

View more >